👓 Omi Pendant + Glasses

Always-on ambient capture. The CI-Omi integration connects the Omi pendant and Omi glasses to your local Companion Hub, logging continuous audio and conversational context directly to your private Digital Memory Server.

Overview

Omi is an open-source wearable — available as a pendant and glasses form factor — that continuously captures ambient audio and conversations. The CI-Omi integration ensures that all captured audio is processed entirely on your Hub using local speech-to-text (Whisper), and stored as searchable memory events on your hardware — not on Omi’s cloud.

The pendant and glasses sync to your phone’s CI Connect app (or directly to your Hub over Wi-Fi) and the transcripts flow immediately into Digital Memory.

Key Features

Continuous ambient recording — configurable push-to-talk or always-on capture modes
On-premise transcription — all audio is transcribed by your local Whisper instance
Speaker diarization — (optional) identifies and labels distinct speakers in multi-person conversations
Conversation summaries — local LLM generates concise summaries of captured sessions
Memory tagging — automatically tags transcripts with location, time, participants, and topics using AI
Privacy-first — raw audio is never retained after transcription (configurable)
Omi glasses + pendant — works with both Omi hardware form factors

Supported Hardware

Device	Connection	Status
Omi Pendant (Gen 1 & 2)	Bluetooth LE → Phone relay	✅
Omi Glasses	Bluetooth LE → Phone relay	✅
Omi (Wi-Fi mode)	Wi-Fi Direct to Hub	✅

Use Cases

Capture every meeting and conversation for private, searchable recall
“What did we discuss about the project deadline last Tuesday?” — answered by your local agent
Build a continuous personal journal of conversations without a cloud subscription
Combine with Mentra Live visual context for richer memory events
Log voice notes hands-free throughout the day

Architecture


Omi Pendant / Glasses
    │  Bluetooth LE
    ▼
CI Connect (mobile relay) OR Wi-Fi Direct
    │
    ▼
CI-Omi Hub App
├── Audio receiver
├── Whisper transcription (local)
├── Diarization engine (optional)
├── LLM summarization
└── Memory bridge
        │
        ▼
Digital Memory Server
(transcripts, summaries, conversation events)

Setup

Install CI-Omi from Hub app store

Search for Omi Pendant + Glasses in Companion Hub and install.

Pair your Omi device

Option A — via phone relay (recommended):

Install CI Connect on your phone and pair it with your Hub
In CI Connect → Wearables → Add Omi
Follow the Bluetooth pairing prompts

Option B — direct Wi-Fi:

Hold the Omi button for 5 seconds to enter Wi-Fi setup mode
Connect to the Omi’s temporary Wi-Fi AP and navigate to 192.168.4.1
Enter your Hub’s Wi-Fi network credentials and Hub address

Configure transcription

Open http://omi.ci.localhost → Settings → Transcription:

Select your Whisper model (larger models are more accurate but slower)
Enable/disable speaker diarization
Set audio language or use auto-detect

Set retention policy

Under Settings → Audio Retention, choose whether to keep raw audio files after transcription. For maximum privacy, enable Transcribe and delete.

Usage

Reviewing Conversations

All transcripts appear in Digital Memory under the Omi source. Each entry includes a full transcript, an AI-generated summary, and a list of topics and participants detected in the session.

Asking About Past Conversations

From the Digital Memory search or a Companion Agent chat:

“Summarize my conversations from today” “What did we decide in the meeting about the product roadmap?”

Starting/Stopping Capture

Pendant: Single press to start capture, single press to stop
Glasses: Tap the frame once to start, once to stop
Always-on mode: Continuous capture with silence detection pausing

Troubleshooting

Omi not connecting via phone relay Ensure Bluetooth is enabled on your phone and within range of the pendant. Check that CI Connect has Bluetooth permission. Restart the Omi by pressing and holding the button for 10 seconds.

Transcripts are empty or inaccurate Verify Whisper is running in Hub (check Hub → Apps → Whisper status). Try a larger Whisper model for better accuracy in noisy environments.

Summaries not generating Check that an LLM service (Ollama, Open WebUI) is running in Hub and is configured in Omi → Settings → Summarization → LLM Endpoint.

High Hub CPU usage Whisper transcription is CPU/GPU intensive. Consider scheduling batch transcription during off-peak hours in Settings → Transcription → Schedule.